Usage

Finding your Cluster ID

Your clusters are available under the Private Clusters page in the Paperspace console. Here you can find information about your cluster, including your cluster ID – you will need to specify this ID in order to designate your private cluster as the place to run experiments, notebooks, etc.

Using Gradient private clusters via the Gradient CLI

gradient <command> ... --clusterId <your-cluster-ID>

--clusterId string Cluster ID for this processing site, e.g. "clxxxxxxx".

You will need to provide your API key to authenticate your requests. Learn how to obtain and set your API key here.

A complete example of utilizing Gradient features on a cluster might look like this:

gradient experiments run singlenode --name experiment1 \
--projectId prgydf45k \
--clusterId cl53waq2x \
--machineType c5.xlarge \
--container tensorflow/tensorflow:1.13.1-py3 \
--experimentEnv "{\"EPOCHS_EVAL\":\"5\",\"TRAIN_EPOCHS\":\"10\",\"MAX_STEPS\":\"1000\",\"EVAL_SECS\":\"10\"}" \
--workspaceUrl https://github.com/Paperspace/mnist-sample.git \
--command "pip install -r requirements.txt && python mnist.py"

In order to run workloads on your Gradient cluster, you must specify the clusterID parameter on most Gradient primary commands, including:

  • experiments

  • deployments

  • jobs

  • models

  • notebooks

  • tensorboards

If you don't supply the clusterId parameter, then your command will default to run on Paperspace instances, which are not part of your private cluster environment.

To avoid having to re-enter the Cluster ID, and if you want the configuration to be reusable and checked into source control, another option is using the Gradient Config File. This file can contain the clusterID parameter in addition to many other common settings.

Using Gradient clusters via the Web UI

When creating a notebook, an experiment, or a model deployment, select your private cluster in the console, then select an instance type that's available in your cluster.